Select: Apply a genetic algorithm to find the optimal co-variates in...

In stat243proj/GA: Appllies a modified genetic algorithm to some dataset, in order to assess what combination of predictor variables are best able to predict

Description

Select() is used to apply the genetic algorithm to some input dataset outputs a list containing a list of the indiivdials in the final generation' of the genetic algorithm, a matrix of the fitness values of all individuals in all generations and the fitted model from glm for the best individual

Usage

Select(dataset, response.name, userfunc = "AIC", user.family = "gaussian",
  flag.log.scale = TRUE, frac.replace = 0.2, Niter = 100, Nruns = 1,
  mutate.rate = FALSE, plot.flag = TRUE)

Arguments

dataset	A matrix, datatable, or dataframe
response.name	The name of the column in dataset that will act as the response variable to be predicted
userfunc	A fitness function that operates on a model that could be provided by the user. The default is the Aikake Information Criteria or "AIC".
user.family	Model family name to be passed to glm. Default is "gaussian"
flag.log.scale	Default is TRUE if the log of the predictor varaible is to be fit.
frac.replace	Fraction of worst parents to be replaced with the best children in each generation
Niter	Maximum number of iterations during each run. Default is 100
Nruns	Number of times genetic algorithm is run. Default is 1
mutate.rate	Genetic algorithm mutation rate. If set to FALSE it is automatically determined. A value of 0.01 is suggested
plot.flag	Set to TRUE to plot the evolution of the population of individuals over the progression of the algorithm

Value

Select() produces a single list each of whose elements are a list or vector of objects partaining to each of the Nruns times that the genetic algorithm was run. Each list element is named according the type of data it contains:

• LastGen is a list of the last generation of solutions produced during each run

• Fitness is a list of the complete fitness matrix representing all generations during each run

• BestModel is a list of the best model produced during each run

• BestFitness is a vector of the fitness of the best model from each run

Examples

baseball = read.table(file.choose(),header=TRUE)
out <- Select(dataset=baseball, response.name="salary", Niter=50, Nruns=1, mutate.rate = 0.01)

stat243proj/GA documentation built on May 15, 2019, 5:03 a.m.